Class JPEGCompression

The JPEG standard defines a flow of level shift, DCT and quantization for forward JPEG transform and inverse level shift, IDCT and de-quantization for inverse JPEG transform. This group has the functions for both forward and inverse functions.

Inheritance

System.Object

JPEGCompression

Implements

System.IDisposable

Inherited Members

System.Object.Equals(System.Object)

System.Object.Equals(System.Object, System.Object)

System.Object.GetHashCode()

System.Object.GetType()

System.Object.MemberwiseClone()

System.Object.ReferenceEquals(System.Object, System.Object)

System.Object.ToString()

Namespace: ManagedCuda.NPP

Assembly: NPP.dll

Syntax

public class JPEGCompression : IDisposable

Constructors

| Improve this Doc View Source

JPEGCompression()

Initializes DCT state structure and allocates additional resources

Declaration

public JPEGCompression()

Methods

| Improve this Doc View Source

DCTQuant16Fwd8x8LS(NPPImage_16sC1, NPPImage_16sC1, NppiSize)

Inverse DCT in WebP decoding. Input is the bitstream that contains the coefficients of 16x16 blocks. These coefficients are based on a 4x4 sub-block unit, e.g., Coeffs in 0th 4x4 block, 1st 4x4 block 2nd 4x4 block, etc. Output is the coefficients after inverse DCT transform. The output is put in an image format (i.e. raster scan order), different from the input order.

Declaration

public void DCTQuant16Fwd8x8LS(NPPImage_16sC1 src, NPPImage_16sC1 dst, NppiSize oSizeRoi)

Parameters

Type	Name	Description
NPPImage_16sC1	src	Source image.
NPPImage_16sC1	dst	Destination image
NppiSize	oSizeRoi	Roi size (in pixels).

| Improve this Doc View Source

DCTQuant16Fwd8x8LS(NPPImage_8uC1, NPPImage_16sC1, NppiSize, CudaDeviceVariable<UInt16>)

Forward DCT, quantization and level shift part of the JPEG encoding, 16-bit short integer. Input is expected in 8x8 macro blocks and output is expected to be in 64x1 macro blocks. The new version of the primitive takes the ROI in image pixel size and works with DCT coefficients that are in zig-zag order.

Declaration

public void DCTQuant16Fwd8x8LS(NPPImage_8uC1 src, NPPImage_16sC1 dst, NppiSize oSizeRoi, CudaDeviceVariable<ushort> pQuantizationTable)

Parameters

Type	Name	Description
NPPImage_8uC1	src	Source image.
NPPImage_16sC1	dst	Destination image
NppiSize	oSizeRoi	Roi size (in pixels).
CudaDeviceVariable<System.UInt16>	pQuantizationTable	Quantization Table in zig-zag order.

| Improve this Doc View Source

DCTQuant16Inv8x8LS(NPPImage_16sC1, NPPImage_8uC1, NppiSize, CudaDeviceVariable<UInt16>)

Inverse DCT, de-quantization and level shift part of the JPEG decoding, 16-bit short integer. Input is expected in 64x1 macro blocks and output is expected to be in 8x8 macro blocks. The new version of the primitive takes the ROI in image pixel size and works with DCT coefficients that are in zig-zag order.

Declaration

public void DCTQuant16Inv8x8LS(NPPImage_16sC1 src, NPPImage_8uC1 dst, NppiSize oSizeRoi, CudaDeviceVariable<ushort> pQuantizationTable)

Parameters

Type	Name	Description
NPPImage_16sC1	src	Source image.
NPPImage_8uC1	dst	Destination image
NppiSize	oSizeRoi	Roi size (in pixels).
CudaDeviceVariable<System.UInt16>	pQuantizationTable	Quantization Table in zig-zag order.

| Improve this Doc View Source

DCTQuantFwd8x8LS(NPPImage_8uC1, NPPImage_16sC1, CudaDeviceVariable<UInt16>, NppiSize)

Forward DCT, quantization and level shift part of the JPEG encoding. Input is expected in 8x8 macro blocks and output is expected to be in 64x1 macro blocks.

Declaration

public static void DCTQuantFwd8x8LS(NPPImage_8uC1 src, NPPImage_16sC1 dst, CudaDeviceVariable<ushort> QuantFwdTable, NppiSize oSizeRoi)

Parameters

Type	Name	Description
NPPImage_8uC1	src	Source image.
NPPImage_16sC1	dst	Destination image
CudaDeviceVariable<System.UInt16>	QuantFwdTable	Forward quantization tables for JPEG encoding created using QuantInvTableInit()
NppiSize	oSizeRoi	Roi size (in macro blocks?).

| Improve this Doc View Source

DCTQuantFwd8x8LS(NPPImage_8uC1, NPPImage_16sC1, NppiSize, CudaDeviceVariable<Byte>)

Forward DCT, quantization and level shift part of the JPEG encoding. Input is expected in 8x8 macro blocks and output is expected to be in 64x1 macro blocks. The new version of the primitive takes the ROI in image pixel size and works with DCT coefficients that are in zig-zag order.

Declaration

public void DCTQuantFwd8x8LS(NPPImage_8uC1 src, NPPImage_16sC1 dst, NppiSize oSizeRoi, CudaDeviceVariable<byte> QuantFwdTable)

Parameters

Type	Name	Description
NPPImage_8uC1	src	Source image.
NPPImage_16sC1	dst	Destination image
NppiSize	oSizeRoi	Roi size (in pixels).
CudaDeviceVariable<System.Byte>	QuantFwdTable	Quantization Table in zig-zag order

| Improve this Doc View Source

DCTQuantInv8x8LS(NPPImage_16sC1, NPPImage_8uC1, CudaDeviceVariable<UInt16>, NppiSize)

Inverse DCT, de-quantization and level shift part of the JPEG decoding. Input is expected in 64x1 macro blocks and output is expected to be in 8x8 macro blocks.

Declaration

public static void DCTQuantInv8x8LS(NPPImage_16sC1 src, NPPImage_8uC1 dst, CudaDeviceVariable<ushort> QuantInvTable, NppiSize oSizeRoi)

Parameters

Type	Name	Description
NPPImage_16sC1	src	Source image.
NPPImage_8uC1	dst	Destination image
CudaDeviceVariable<System.UInt16>	QuantInvTable	Inverse quantization tables for JPEG decoding created using QuantInvTableInit()
NppiSize	oSizeRoi	Roi size (in macro blocks?).

| Improve this Doc View Source

DCTQuantInv8x8LS(NPPImage_16sC1, NPPImage_8uC1, NppiSize, CudaDeviceVariable<Byte>)

Inverse DCT, de-quantization and level shift part of the JPEG decoding. Input is expected in 64x1 macro blocks and output is expected to be in 8x8 macro blocks. The new version of the primitive takes the ROI in image pixel size and works with DCT coefficients that are in zig-zag order.

Declaration

public void DCTQuantInv8x8LS(NPPImage_16sC1 src, NPPImage_8uC1 dst, NppiSize oSizeRoi, CudaDeviceVariable<byte> QuantInvTable)

Parameters

Type	Name	Description
NPPImage_16sC1	src	Source image.
NPPImage_8uC1	dst	Destination image
NppiSize	oSizeRoi	Roi size (in pixels).
CudaDeviceVariable<System.Byte>	QuantInvTable	Quantization Table in zig-zag order.

| Improve this Doc View Source

DecodeHuffmanScanHost(Byte[], Int32, Int32, Int32, Int32, Int32, Int16[], Int16[], Int16[], Int32[], NppiDecodeHuffmanSpec[], NppiDecodeHuffmanSpec[], NppiSize[])

Huffman Decoding of the JPEG decoding on the host.

Input is expected in byte stuffed huffman encoded JPEG scan and output is expected to be 64x1 macro blocks.

Declaration

public static void DecodeHuffmanScanHost(byte[] pSrc, int restartInterval, int Ss, int Se, int Ah, int Al, short[] pDstY, short[] pDstCb, short[] pDstCr, int[] nDstStep, NppiDecodeHuffmanSpec[] pHuffmanTableDC, NppiDecodeHuffmanSpec[] pHuffmanTableAC, NppiSize[] oSizeROI)

Parameters

Type	Name	Description
System.Byte[]	pSrc	Byte-stuffed huffman encoded JPEG scan.
System.Int32	restartInterval	Restart Interval, see JPEG standard.
System.Int32	Ss	Start Coefficient, see JPEG standard.
System.Int32	Se	End Coefficient, see JPEG standard.
System.Int32	Ah	Bit Approximation High, see JPEG standard.
System.Int32	Al	Bit Approximation Low, see JPEG standard.
System.Int16[]	pDstY	Destination first image channel
System.Int16[]	pDstCb	Destination second image channel
System.Int16[]	pDstCr	Destination third image channel
System.Int32[]	nDstStep	destination image line step.
NppiDecodeHuffmanSpec[]	pHuffmanTableDC	DC Huffman table.
NppiDecodeHuffmanSpec[]	pHuffmanTableAC	AC Huffman table.
NppiSize[]	oSizeROI	ROI

| Improve this Doc View Source

DecodeHuffmanScanHost(Byte[], Int32, Int32, Int32, Int32, Int32, Int16[], Int32, NppiDecodeHuffmanSpec, NppiDecodeHuffmanSpec, NppiSize)

Huffman Decoding of the JPEG decoding on the host.

Input is expected in byte stuffed huffman encoded JPEG scan and output is expected to be 64x1 macro blocks.

Declaration

public static void DecodeHuffmanScanHost(byte[] pSrc, int restartInterval, int Ss, int Se, int Ah, int Al, short[] pDst, int nDstStep, NppiDecodeHuffmanSpec pHuffmanTableDC, NppiDecodeHuffmanSpec pHuffmanTableAC, NppiSize oSizeROI)

Parameters

Type	Name	Description
System.Byte[]	pSrc	Byte-stuffed huffman encoded JPEG scan.
System.Int32	restartInterval	Restart Interval, see JPEG standard.
System.Int32	Ss	Start Coefficient, see JPEG standard.
System.Int32	Se	End Coefficient, see JPEG standard.
System.Int32	Ah	Bit Approximation High, see JPEG standard.
System.Int32	Al	Bit Approximation Low, see JPEG standard.
System.Int16[]	pDst	Destination image pointer
System.Int32	nDstStep	destination image line step.
NppiDecodeHuffmanSpec	pHuffmanTableDC	DC Huffman table.
NppiDecodeHuffmanSpec	pHuffmanTableAC	AC Huffman table.
NppiSize	oSizeROI	ROI

| Improve this Doc View Source

DecodeHuffmanSpecFreeHost(NppiDecodeHuffmanSpec)

Frees the host memory allocated by nppiDecodeHuffmanSpecInitAllocHost_JPEG.

Declaration

public static void DecodeHuffmanSpecFreeHost(NppiDecodeHuffmanSpec pHuffmanSpec)

Parameters

Type	Name	Description
NppiDecodeHuffmanSpec	pHuffmanSpec	Pointer to the Huffman table for the decoder

| Improve this Doc View Source

DecodeHuffmanSpecGetBufSize()

Returns the length of the NppiDecodeHuffmanSpec structure.

Declaration

public static int DecodeHuffmanSpecGetBufSize()

Returns

Type	Description
System.Int32	the length of the NppiDecodeHuffmanSpec structure.

| Improve this Doc View Source

DecodeHuffmanSpecInitAllocHost(Byte[], NppiHuffmanTableType)

Allocates memory and creates a Huffman table in a format that is suitable for the decoder on the host.

Declaration

public static NppiDecodeHuffmanSpec DecodeHuffmanSpecInitAllocHost(byte[] pRawHuffmanTable, NppiHuffmanTableType eTableType)

Parameters

Type	Name	Description
System.Byte[]	pRawHuffmanTable	Huffman table formated as specified in the JPEG standard.
NppiHuffmanTableType	eTableType	Enum specifying type of table (nppiDCTable or nppiACTable).

Returns

Type	Description
NppiDecodeHuffmanSpec	Huffman table for the decoder

| Improve this Doc View Source

DecodeHuffmanSpecInitHost(Byte[], NppiHuffmanTableType, NppiDecodeHuffmanSpec)

Creates a Huffman table in a format that is suitable for the decoder on the host.

Declaration

public static void DecodeHuffmanSpecInitHost(byte[] pRawHuffmanTable, NppiHuffmanTableType eTableType, NppiDecodeHuffmanSpec pHuffmanSpec)

Parameters

Type	Name	Description
System.Byte[]	pRawHuffmanTable	Huffman table formated as specified in the JPEG standard.
NppiHuffmanTableType	eTableType	Enum specifying type of table (nppiDCTable or nppiACTable)
NppiDecodeHuffmanSpec	pHuffmanSpec	Pointer to the Huffman table for the decoder

| Improve this Doc View Source

DecodeJobCreateFinalize(ref NppiJpegDecodeJob)

Initializes a job that has to be called at the end of decoding, in order to convert temporary representation of DCT coefficients to the final one.

Declaration

public void DecodeJobCreateFinalize(ref NppiJpegDecodeJob pJob)

Parameters

Type	Name	Description
NppiJpegDecodeJob	pJob	pJob.pFrame should point to valid frame description. pJob.pScan will be overwritten.

| Improve this Doc View Source

Dispose()

Dispose

Declaration

public virtual void Dispose()

| Improve this Doc View Source

Dispose(Boolean)

For IDisposable

Declaration

protected virtual void Dispose(bool fDisposing)

Parameters

Type	Name	Description
System.Boolean	fDisposing

| Improve this Doc View Source

EncodeHuffmanGetSize(NppiSize, Int32)

Calculates the size of the temporary buffer for huffman encoding.

Declaration

public static int EncodeHuffmanGetSize(NppiSize oSize, int nChannels)

Parameters

Type	Name	Description
NppiSize	oSize	Image Dimension
System.Int32	nChannels	Number of channels

Returns

Type	Description
System.Int32	the size of the temporary buffer

| Improve this Doc View Source

EncodeHuffmanScan(NPPImage_16sC1, Int32, Int32, Int32, Int32, Int32, CudaDeviceVariable<Byte>, ref Int32, NppiEncodeHuffmanSpec, NppiEncodeHuffmanSpec, NppiSize, CudaDeviceVariable<Byte>)

Huffman Encoding of the JPEG Encoding.

Input is expected to be 64x1 macro blocks and output is expected as byte stuffed huffman encoded JPEG scan.

Declaration

public static void EncodeHuffmanScan(NPPImage_16sC1 pSrc, int restartInterval, int Ss, int Se, int Ah, int Al, CudaDeviceVariable<byte> pDst, ref int nLength, NppiEncodeHuffmanSpec pHuffmanTableDC, NppiEncodeHuffmanSpec pHuffmanTableAC, NppiSize oSizeROI, CudaDeviceVariable<byte> buffer)

Parameters

Type	Name	Description
NPPImage_16sC1	pSrc	Source image.
System.Int32	restartInterval	Restart Interval, see JPEG standard.
System.Int32	Ss	Start Coefficient, see JPEG standard.
System.Int32	Se	End Coefficient, see JPEG standard.
System.Int32	Ah	Bit Approximation High, see JPEG standard.
System.Int32	Al	Bit Approximation Low, see JPEG standard.
CudaDeviceVariable<System.Byte>	pDst	Byte-stuffed huffman encoded JPEG scan.
System.Int32	nLength	Byte length of the huffman encoded JPEG scan.
NppiEncodeHuffmanSpec	pHuffmanTableDC	DC Huffman table.
NppiEncodeHuffmanSpec	pHuffmanTableAC	AC Huffman table.
NppiSize	oSizeROI	ROI
CudaDeviceVariable<System.Byte>	buffer	Scratch buffer

| Improve this Doc View Source

EncodeHuffmanScan(NPPImage_16sC1[], Int32, Int32, Int32, Int32, Int32, CudaDeviceVariable<Byte>, ref Int32, NppiEncodeHuffmanSpec[], NppiEncodeHuffmanSpec[], NppiSize[], CudaDeviceVariable<Byte>)

Huffman Encoding of the JPEG Encoding.

Input is expected to be 64x1 macro blocks and output is expected as byte stuffed huffman encoded JPEG scan.

Declaration

public static void EncodeHuffmanScan(NPPImage_16sC1[] pSrc, int restartInterval, int Ss, int Se, int Ah, int Al, CudaDeviceVariable<byte> pDst, ref int nLength, NppiEncodeHuffmanSpec[] pHuffmanTableDC, NppiEncodeHuffmanSpec[] pHuffmanTableAC, NppiSize[] oSizeROI, CudaDeviceVariable<byte> buffer)

Parameters

Type	Name	Description
NPPImage_16sC1[]	pSrc	Source image.
System.Int32	restartInterval	Restart Interval, see JPEG standard.
System.Int32	Ss	Start Coefficient, see JPEG standard.
System.Int32	Se	End Coefficient, see JPEG standard.
System.Int32	Ah	Bit Approximation High, see JPEG standard.
System.Int32	Al	Bit Approximation Low, see JPEG standard.
CudaDeviceVariable<System.Byte>	pDst	Byte-stuffed huffman encoded JPEG scan.
System.Int32	nLength	Byte length of the huffman encoded JPEG scan.
NppiEncodeHuffmanSpec[]	pHuffmanTableDC	DC Huffman table.
NppiEncodeHuffmanSpec[]	pHuffmanTableAC	AC Huffman table.
NppiSize[]	oSizeROI	ROI
CudaDeviceVariable<System.Byte>	buffer	Scratch buffer

| Improve this Doc View Source

EncodeHuffmanSpecFree(NppiEncodeHuffmanSpec)

Frees the memory allocated by nppiEncodeHuffmanSpecInitAlloc_JPEG.

Declaration

public static void EncodeHuffmanSpecFree(NppiEncodeHuffmanSpec pHuffmanSpec)

Parameters

Type	Name	Description
NppiEncodeHuffmanSpec	pHuffmanSpec	Pointer to the Huffman table for the encoder

| Improve this Doc View Source

EncodeHuffmanSpecGetBufSize()

Returns the length of the NppiEncodeHuffmanSpec structure.

Declaration

public static int EncodeHuffmanSpecGetBufSize()

Returns

Type	Description
System.Int32	length of the NppiEncodeHuffmanSpec structure.

| Improve this Doc View Source

EncodeHuffmanSpecInit(Byte[], NppiHuffmanTableType, NppiEncodeHuffmanSpec)

Creates a Huffman table in a format that is suitable for the encoder.

Declaration

public static void EncodeHuffmanSpecInit(byte[] pRawHuffmanTable, NppiHuffmanTableType eTableType, NppiEncodeHuffmanSpec pHuffmanSpec)

Parameters

Type	Name	Description
System.Byte[]	pRawHuffmanTable	Huffman table formated as specified in the JPEG standard.
NppiHuffmanTableType	eTableType	Enum specifying type of table (nppiDCTable or nppiACTable).
NppiEncodeHuffmanSpec	pHuffmanSpec	Pointer to the Huffman table for the decoder

| Improve this Doc View Source

EncodeHuffmanSpecInitAlloc(Byte[], NppiHuffmanTableType)

Allocates memory and creates a Huffman table in a format that is suitable for the encoder.

Declaration

public static NppiEncodeHuffmanSpec EncodeHuffmanSpecInitAlloc(byte[] pRawHuffmanTable, NppiHuffmanTableType eTableType)

Parameters

Type	Name	Description
System.Byte[]	pRawHuffmanTable	Huffman table formated as specified in the JPEG standard.
NppiHuffmanTableType	eTableType	Enum specifying type of table (nppiDCTable or nppiACTable).

Returns

Type	Description
NppiEncodeHuffmanSpec	Huffman table for the encoder.

| Improve this Doc View Source

GetDCTBufferSize(NppiSize)

Returns how much memory has to be allocated for DCT coefficient buffers declared in \ref NppiJpegDecodeJobMemory. The returned value may be bigger than simply number of blocks /// 64 /// sizeof (short), because decoder may use slightly bigger temporary representation of data.

Declaration

public SizeT GetDCTBufferSize(NppiSize oBlocks)

Parameters

Type	Name	Description
NppiSize	oBlocks	Size of the interleaved component in blocks.

Returns

Type	Description
SizeT

| Improve this Doc View Source

GetScanDeadzoneSize()

This function returns how much additional memory has to be available after the end of compressed scan data. The following buffers: pCpuScan and pGpuScan in \ref NppiJpegDecodeJobMemory should have size at least pScan->length + nppiJpegDecodeGetScanDeadzoneSize(). The additional memory is needed because the decoder may perform some speculative reads after the end of compressed scan data.

Declaration

public SizeT GetScanDeadzoneSize()

Returns

Type	Description
SizeT

| Improve this Doc View Source

JobCreateMemzero()

Initializes a job that has to be called at the beginning of decoding.

Declaration

public NppiJpegDecodeJob JobCreateMemzero()

Returns

Type	Description
NppiJpegDecodeJob

| Improve this Doc View Source

JobMemorySize(NppiJpegDecodeJob)

Calculates sizes of additional buffers used by the job.

Declaration

public SizeT JobMemorySize(NppiJpegDecodeJob pJob)

Parameters

Type	Name	Description
NppiJpegDecodeJob	pJob	has to point to properly initialized job

Returns

Type	Description
SizeT

| Improve this Doc View Source

JpegDecodeJob(ref NppiJpegDecodeJob, ref NppiJpegDecodeJobMemory)

Executes a job -- part of decoding.

Declaration

public void JpegDecodeJob(ref NppiJpegDecodeJob pJob, ref NppiJpegDecodeJobMemory pMemory)

Parameters

Type	Name	Description
NppiJpegDecodeJob	pJob	has to be initialized by \ref nppiJpegDecodeJobCreateMemzero
NppiJpegDecodeJobMemory	pMemory	has to point to valid structure, except for MEMZERO and FINALIZE

| Improve this Doc View Source

QuantFwdRawTableInit(Byte[], Int32)

Apply quality factor to raw 8-bit quantization table.

This is effectively and in-place method that modifies a given raw quantization table based on a quality factor.

Note that this method is a host method and that the pointer to the raw quantization table is a host pointer.

Declaration

public static void QuantFwdRawTableInit(byte[] QuantRawTable, int nQualityFactor)

Parameters

Type	Name	Description
System.Byte[]	QuantRawTable	Raw quantization table.
System.Int32	nQualityFactor	Quality factor for the table. Range is [1:100].

| Improve this Doc View Source

QuantFwdTableInit(Byte[], UInt16[])

Initializes a quantization table for DCTQuantFwd8x8LS().

The method creates a 16-bit version of the raw table and converts the data order from zigzag layout to original row-order layout since raw quantization tables are typically stored in zigzag format.

This method is a host method. It consumes and produces host data. I.e. the pointers passed to this function must be host pointers. The resulting table needs to be transferred to device memory in order to be used with nppiDCTQuantFwd8x8LS() function.

Declaration

public static void QuantFwdTableInit(byte[] QuantRawTable, ushort[] QuantFwdRawTable)

Parameters

Type	Name	Description
System.Byte[]	QuantRawTable	Host pointer to raw quantization table as returned by QuantFwdRawTableInit(). The raw quantization table is assumed to be in zigzag order.
System.UInt16[]	QuantFwdRawTable	Forward quantization table for use with DCTQuantFwd8x8LS().

| Improve this Doc View Source

QuantInvTableInit(Byte[], UInt16[])

Initializes a quantization table for DCTQuantInv8x8LS().

The DCTQuantFwd8x8LS() method uses a quantization table in a 16-bit format allowing for faster processing. In addition it converts the data order from zigzag layout to original row-order layout. Typically raw quantization tables are stored in zigzag format.

Declaration

public static void QuantInvTableInit(byte[] QuantRawTable, ushort[] QuantInvRawTable)

Parameters

Type	Name	Description
System.Byte[]	QuantRawTable	Raw quantization table.
System.UInt16[]	QuantInvRawTable	Inverse quantization table.

Implements

System.IDisposable